Post on 17-Jun-2020
arX
iv:a
stro
-ph/
0410
536v
1 2
2 O
ct 2
004
1
Accreting neutron stars and black holes:
a decade of discoveriesDIMITRIOS PSALTISDepartment of Physics, University of Arizona, Tucson, AZ 85721 USA
1.1 Introduction
Since their discovery in 1962 (Giacconi et al. 1962), accreting compact ob-
jects in the galaxy have offered unique insights into the astrophysics of the end stagesof stellar evolution and the physics of matter at extreme physical conditions. During
the first three decades of exploration, new phenomena were discovered and under-
stood, such as the periodic pulsations in the X-ray lightcurve of spinning neutron
stars (Giacconi et al. 1971) and the thermonuclear flashes on neutron-star surfaces
that are detected as powerful X-ray bursts (see, e.g., Grindlay et al. 1976; §3).Moreover, the masses of the compact objects were measured in a number of systems,
providing the strongest evidence for the existence of black holes in the universe (Mc-
Clintock & Remillard 1986; §4).
During the last ten years, the launch of X-ray telescopes with unprecedented ca-pabilities, such as RXTE, BeppoSAX, the Chandra X-ray Observatory, and XMM-
Newton opened new windows onto the properties of accreting compact objects.
Examples include the rapid variability phenomena that occur at the dynamical
timescales just outside the neutron-star surfaces and the black-hole horizons (van
der Klis et al. 1996; Strohmayer et al. 1996; §2, §4) as well as atomic lines that havebeen red- and blue-shifted by general relativistic effects in the vicinities of compact
objects (Miller et al. 2002b; Cottam et al. 2001). Accreting neutrons stars and black
holes have been monitored in broad spectral bands, from the radio to γ-rays, leading
to the discovery of highly relativistic jets (Mirabel & Rodriguez 1994; §9), to theindirect imaging of the accretion flows (Horne 1985; §5), and to the possible identifi-
cation of neutron stars with masses close to the maximum value allowed by general
relativity (Barziv et al. 2000). Finally, the theoretical modeling of accretion flows
also experienced significant advances, such as the identification of a whole suit of
stable solution for accretion flows beyond the standard model of geometrically thinaccretion disks (e.g., Narayan & Yi 1994) and of the most promising avenue towards
explaining the very efficient transport of angular momentum in accretion flows (e.g.,
Balbus & Hawley 1991).
The aim of this chapter is to provide a general overview of these recent advancesin the astrophysics of X-ray binaries in our galaxy. The basic concepts have been
reviewed in a number of textbooks (e.g., Shapiro & Teukolsky 1983; Glendenning
2003) and review articles (e.g., White, Nagase, & Parmar 1995) and will only be
briefly mentioned here. Several other classes of compact stellar X-ray sources that
1
2 Accreting neutron stars and black holes:a decade of discoveries
do not involve accretion onto a neutron star or black hole will also not be discussed
in this chapter but are reviewed elsewhere in this volume. These systems include:Isolated neutron stars (§8); Cataclysmic variables (CVs; §10); Supersoft sources (SSS;
§11); Soft Gamma-ray Repeaters and Anomalous X-ray Pulsars (SGRs and AXPs;
§16); and Gamma ray bursts (GRBs; §15). Finally, accreting compact objects in
other galaxies will be reviewed in §11.
1.1.1 X-ray binary systems
Whether a compact object in a binary system is accreting mass in a stable
long-lived phase or not depends mostly on the mode of mass transfer, the ratio of
the mass of the compact object to that of the companion star, and their orbitalseparation. For example, in the case of a neutron star (with a mass ∼ 1.4−2.0 M⊙),
stable mass transfer through the inner Lagrangian point occurs only when the com-
panion fills its Roche lobe and has a mass smaller than that of the neutron star.
In such systems, mass is driven by angular momentum loses due to gravitationalradiation (for very small masses and orbital separations) and magnetic breaking (for
orbital periods ≤ 2 day) or by the evolution of the companion star (for orbital pe-
riods ≥ 2 day). These sources are significantly brighter in the X-rays than in the
optical wavelengths, with the flux at the latter spectral band being mostly due to
reprocessing of the X-ray flux from the outer accretion flows. Binary systems withlow-mass companions to the neutron stars or black holes are called Low-Mass X-ray
Binaries (LMXBs).
A compact object can also accrete matter from a companion star that does not fill
its Roche lobe, if the latter star is losing mass in the form of a stellar wind. For thisprocess to result in a compact star that is a bright X-ray source, the companion star
has to be massive (≥ 10 M⊙) in order to drive a strong wind. In this configuration,
the optical luminosity of the companion star dominates the total emission from the
system and the rate of mass transfer is determined by the strength and speed of the
wind and the orbital separation. Such systems are called High-Mass X-ray Binaries(HMXBs).
The large difference in the companion masses between low- and high-mass X-ray
binaries leads to a number of additional differences between these two classes of
systems. The lifetimes of HMXBs are determined by the evolution of the high-masscompanions and are short (∼ 105 − 107 yr), whereas the lifetimes of the LMXBs are
determined by the mass-transfer process and are longer (∼ 107 − 109 yr). For this
reason, HMXBs are distributed along the galactic plane, as young stellar populations
do, whereas LMXBs are found mostly towards the galactic center and in globular
clusters (Fig. 1.1). Moreover, because neutron stars in HMXBs accrete for a relativelyshort period of time, their magnetic fields do not evolve away from their high birth
values, and hence these neutron stars appear mostly as accretion-powered pulsars.
On the other hand, the prolonged phase of accretion onto neutron stars in LMXBs
is believed to be responsible for the suppression of the stellar fields and the absenceof periodic pulsations in all but a handful of them.
Finally, in LMXBs, the very small sizes of the companion stars can lead to a
number of interesting configurations in systems that are viewed nearly edge on. For
example, in the Accretion Disk Corona (ADC) sources, the X-rays from the central
1.1 Introduction 3
−90 0 90180
−60
−30
0
30
60
Fig. 1.1. Distribution of Low-Mass X-ray Binaries (open symbols) and High-Mass X-rayBinaries (filled symbols) in galactic coordinates (Grimm, Gilfanov & Sunyaev 2002).
objects are scattered towards the observer by electrons in a hot corona that has asize larger than that of the companion, e.g., smoothing out the lightcurves of the
X-ray eclipses (White & Holt 1982). On the other hand, in the so-called dippers, the
shallow X-ray eclipses may not be caused by the companion stars but rather by the
stream of mass transfer from the companion stars to the accretion disks (see, e.g.,White & Swank 1982).
Overall, there are believed to be only a few hundred accreting high-mass and low-mass X-ray binaries in the whole galaxy. Consequently, these binaries are extremely
rare among stellar systems. This is in accord with the large number of improbable
evolutionary steps a primordial binary needs to follow in order to become an X-ray
source with an accreting compact object. Indeed, the progenitors of the compactobjects are believed to be too large to fit in the tight orbits of most X-ray bina-
ries. Moreover, the supernova explosions that precede the formation of the compact
objects may disrupt most systems at the phase prior to the formation of the X-ray
binary. The resolutions to these and other puzzles on the formation and evolution of
X-ray binaries involve exotic and poorly understood binary-evolution processes suchas common-envelope evolution of binary stars (Taam & Sandquist 2000), asymmetric
supernova explosions that impart recoil velocities to the newborn compact objects,
and two- and three-star interactions in the dense stellar fields of globular clusters
(see §8) . The processes that lead to the formation and evolution of X-ray binariesare reviewed in detail in §16.
1.1.2 Accretion onto Compact Objects
An X-ray binary is formed when either the companion star transfers matter
onto the compact object through the inner Lagrangian point or the compact object
captures mass from the wind of the companion star. In both cases, the fate of the
transferred mass depends on the amount of angular momentum it possesses, on the
4 Accreting neutron stars and black holes:a decade of discoveries
physical processes by which it looses angular momentum, and, most importantly,
on the radiation processes by which it cools (see Frank, King & Raine 2002 for acomprehensive review of this subject).
Beginning in the early 1970’s and for the next two decades, most of the model-
ing effort of accretion flows onto neutron stars and black holes was based on two
restrictive assumptions. First, accretion flows were assumed to be loosing angularmomentum at high rates because of an unspecified process, with the effective kine-
matic viscosity typically taken to be proportional to the pressure (see, e.g., Shakura
& Sunyaev 1973; these solutions are often called α-disks, named after the constant
of proportionality). Second, radiation processes were assumed to be very efficient, so
that the resulting accretion flows were relatively cool, in the form of geometricallythin accretion disks. The first of these assumptions stemmed from calculations that
showed the inefficiency of microscopic viscosity to account for the high inferred rates
of mass accretion in the observed sources (see Pringle 1981 for a review). The second
assumption, on the other hand, was relaxed in a number of studies (e.g., Shapiro,Lightman & Eardley 1976) but the resulting solutions were shown to be unstable
(e.g., Piran 1978).
During the last decade, theoretical models of accretion flows onto compact objects
became increasingly more sophisticated and diverse because of two major develop-
ments. First was the identification of a magnetohydrodynamic instability in differ-entially rotating flows (the magneto-rotational instability, or MRI; Balbus & Hawley
1991, 1998), which allows seed magnetic fields of infinitesimal strength in the flow
to get enhanced and tangled. This was shown to lead to a fully-developed magneto-
hydrodynamic turbulence and provide an efficient mechanism of angular momentumtransport, as envisioned in the earlier empirical models (Balbus & Papaloizou 1999).
Studies of the non-linear development of the instability, its level of saturation (e.g.,
Sano et al. 2004), as well as of its effect on the overall properties of accretion disks
(e.g., Hawley & Krolik 2001, 2002; Armitage et al. 2001; McKinney & Gammie 2002)
require large-scale numerical simulations and are all subjects of intense research ef-forts.
The second development is related to the discovery of new stable solutions to the
hydrodynamic equations that describe radiatively inefficient accretion flows (e.g.,
Narayan & Yi 1994). In these solutions, the electrons and ions have different andhigh temperatures, the accretion flows are geometrically thick, and most of their
potential energy is not radiated away but is rather advected towards the compact
objects; these are the so-called Advection-Dominated Accretion Flows (ADAFs).
Besides being interesting new theoretical solutions to the hydrodynamics equations,
advection-dominated flows provided a framework within which the anomalously lowefficiency of accretion onto several black-holes in centers of galaxies (Narayan et al.
1995) and in the quiescent states of X-ray transients (Narayan et al. 1996) can be
understood.
Recently, a number of basic properties of these advection-dominated solutions havebeen scrutinized. The basic assumption that electrons and ions are coupled ineffi-
ciently, mostly due to Coulomb scattering, has been revised, taking into account the
effects of magnetic fields (e.g., Quataert & Gruzinov 1999). Advection-dominated
flows were shown to be capable of launching strong outflows (Advection-Dominated
1.2 Pulsating neutron stars 5
Fig. 1.2. Results of a numerical simulation of a magneto-hydrodynamic accretion flow ontoa black hole. The panels show the logarithm of the density, specific entropy, square of thetoroidal magnetic field, and the r − φ component of the Maxwell stress tensor (Stone &Pringle 2001).
Inflow/Outflow Solutions or ADIOS; Blandford & Begelman 1999). Moreover, for
a wide area of the parameter space, numerical (Stone et al. 1999; Igumenshchev etal. 2000) and analytical studies (Narayan et al. 2000) showed that the solutions are
convectively unstable (Convection-Dominated Accretion Flows or CDAFs). Finally,
the effects of the magneto-rotational instability on the properties of radiatively ineffi-
cient flows have also been investigated recently both in the Newtonian (see, Fig. 1.2;Stone & Pringle 2001; Igumenshchev et al. 2003) and in the general relativistic
regimes (DeVilliers et al. 2003; Gammie et al. 2003).
The final ingredient in the models of accretion flows onto compact objects is the
interaction of the flows with the objects themselves. This is the region in the accretion
flows where most of the high-energy radiation is produced and hence is the one that
is probed by observations with X-ray and γ-ray telescopes. Clearly, the interaction
depends on whether the compact object is a black hole or a neutron star, and inthe latter case, on whether the neutron star is strongly or weakly magnetic. The
main observational manifestation of these differences is the presence or absence of
pulsations in the X-ray lightcurves of the systems, which reflects the strength of the
magnetic fields of the central objects. In the rest of this chapter the observationalproperties of the pulsating and non-pulsating X-ray binaries, as well as the current
efforts for their theoretical modeling, will be reviewed.
1.2 Pulsating neutron stars
Neutron stars possess some of the strongest magnetic fields observed in na-
ture. The origin of these magnetic fields is only poorly understood, mostly due to
our inability to observe directly the magnetic fields of the cores of pre-supernova
stars, which collapse to form the neutron stars, and model their amplification. How-
6 Accreting neutron stars and black holes:a decade of discoveries
ever, observations of isolated radio pulsars (§7) and magnetars (§14) provide strong
evidence that neutron-star magnetic fields range between ∼ 108 G and ∼ 1015 G.When a strongly magnetic neutron star accretes plasma from a companion star or
the interstellar medium, its magnetic field becomes dynamically important close to
the stellar surface and determines the properties of the accretion flow. The radius
at which the effects of the magnetic field dominate all others is called the Alfvenradius and its precise definition depends on the mode of accretion (i.e., thin-disk
vs. quasi-radial), the topology of the magnetic field (i.e., dipolar vs. multipolar), etc.
For thin-disk accretion onto a neutron star, the Alfven radius is defined as the radius
at which magnetic stresses remove efficiently the angular momentum of the accreting
material (see Ghosh & Lamb 1991 and references therein). For a surface magneticfield strength of 1012 G and a mass accretion rate comparable to the Eddington
critical rate, the Alfven radius is of order ∼ 100 neutron-star radii.
The fate of the accreting material after it interacts with the stellar magnetic field
near the Alfven radius depends on the spin frequency of the neutron star. If the stellarspin frequency is smaller than the orbital frequency of matter at the interaction
radius, then the accreting material is forced into corotation with the star and is
channeled along field lines onto the magnetic poles. As the neutron star spins and
the observer sees a different aspect of the hotter magnetic poles, the X-ray flux
received is modulated at the stellar spin frequency and an accretion-powered pulsaris produced. On the other hand, if the stellar spin frequency is larger than the orbital
frequency of matter at the interaction radius, then the material cannot overcome the
centrifugal barrier in order to accrete onto the star. The fate of matter in this case
is presently unknown, but it is often assumed that matter eventually escapes theneutron star in the form of a wind. Magnetic neutron stars in this configuration are
often said to be in the propeller regime (after Illarionov & Sunyaev 1975).
The neutron star itself also reacts differently to the accretion of matter depending
on its magnetic field strength, its spin frequency, and the mass accretion rate. Mag-
netic field lines rotate at the spin frequency of the star and couple the stellar surfaceto the accreting material. As a result, they transfer angular momentum from the
accreting material to the neutron star, if the former is spinning faster than the latter
or from the neutron star to the accreting material, in the opposite situation. Both
situations occur simultaneously in an accreting system, since the orbital frequencyof matter decreases with increasing radius. The overall effect is a net torque on the
neutron star, which can be either positive (spin-up) or negative (spin-down). The
magnitude of the torque on the star is expected to increase with increasing mass
accretion rate and with increasing magnetic field strength (see Ghosh & Lamb 1979;
Ghosh & Lamb 1991). Clearly, for every magnetic field strength and mass accretionrate, there is a critical spin frequency at which the net torque on the star is zero. This
frequency corresponds to an equilibrium, towards which the neutron star evolves in
its lifetime. For a surface, dipolar magnetic field with a strength of 1012 G and a
mass accretion rate comparable to the Eddington critical rate, the equilibrium spinfrequency is of order of a few tenths of a Hertz (Ghosh & Lamb 1992).
Accretion-powered pulsars provide currently the best systems in which the spin
frequencies and magnetic field strengths of accreting neutron stars can be studied.
Two distinct classes of such pulsars are known: pulsars with periods of order a second,
1.2 Pulsating neutron stars 7
Fig. 1.3. The spin and orbital periods of classical accretion-powered pulsars (the Corbetdiagram; after Bildsten et al. 1997).
which are found mostly in high-mass X-ray binaries, and pulsars with millisecond
spin periods, which are found in binary systems with very short orbital periods (see§1.1.2).
1.2.1 Classical (slow) accretion-powered pulsars
The detection of coherent pulsations from an accreting X-ray source, in1971 (Giacconi et al. 1971), provided the strongest evidence, at the time, that the
compact objects in many of these sources were neutron stars. Since then, accretion-
powered pulsars with periods of the order of one second or more have been studied
extensively with every X-ray satellite. In recent years, the long-term monitoring ofsuch pulsars with BATSE as well as the detailed spectral studies with RXTE and
BeppoSAX provided a unique look into the properties of these systems, resolving
some long-standing questions and posing a number of new ones (see Bildsten et
al. 1997 and Heindl et al. 2004 for comprehensive reviews of the accretion-powered
pulsars discussed in this section).
The vast majority of slow accretion-powered pulsars are found in high-mass X-raybinaries; only five of them (Her X-1, 4U 1626−67, GX 1+4, GRO J1744−28, and
2A 1822−371) have low-mass companions. Indeed, most low-mass X-ray binaries are
old systems and their prolonged phase of accretion is thought to have suppressed the
magnetic fields of the neutron stars and to have spun them up to millisecond periods(see §1.2.2). On the other hand, high-mass X-ray binaries are younger systems and
the neutron stars in them are expected to have magnetic fields that are dynamically
important.
The properties of the high-mass binary systems in which slow pulsars reside can
be described more easily on the diagram that correlates their spin to their orbital
8 Accreting neutron stars and black holes:a decade of discoveries
Fig. 1.4. (Left) The dependence of the spin-up rate on the pulsed flux for the sourceA 0535+262, as observed by BATSE; the dashed and dotted curves show the best-fit lineand the theoretical prediction, respectively (see text). (Right) The evolution of the spinfrequency of Cen X-3 as observed by BATSE (Bildsten et al. 1997).
periods (the Corbet diagram; Fig. 1.3). About half of the slow pulsars are orbit-
ing main-sequence Be stars, whereas the remaining pulsars are orbiting evolved OB
supergiants. The systems with Be companions are generally eccentric transient sys-tems, in which the companion stars are not filling their Roche lobes and the pulsars
become detectable during periastron passages (§5). The properties of the systems
with supergiant companions depend on whether these stars fill their Roche lobes or
not. If they do, matter is transfered onto the neutron stars via the inner Lagrangianpoint of their binary potential possessing significant angular momentum and forming
a geometrically thin accretion disk. On the other hand, for companion stars that do
not fill their Roche lobes, mass lost via a radiation-driven wind is captured by the
neutron star at rates that are typically lower than the disk-fed systems.
1.2.1.1 Spin-period evolution
The mode of mass transfer onto the neutron stars, which depends on the
properties of the binary systems, also determines the spin evolution of the neutron
stars. The long-term aspect of this dependence is clearly visible in Fig. 1.3. The
systems with Roche-lobe filling supergiants have short spin periods that are anti-correlated with the orbital periods; the systems with underfilling supergiants have
long spin periods that do not show any correlation with orbital periods; and the Be
transient systems have long orbital periods which are positively correlated to the
orbital periods. These correlations are believed to depend strongly on the mode and
efficiency of mass transfer from the companion stars to the neutron stars but areonly poorly understood (see, e.g., Waters & van Kerkwijk 1989).
A clear look into the short-term dependence of the spin periods of neutron stars
on the properties of the accretion flows was made possible because of the intense
monitoring of several accretion-powered pulsars with the BATSE experiment onboardCGRO. Contrary to earlier results, the measurements with BATSE revealed that
transient and persistent sources show two different types of spin-period evolution
(Bildsten et al. 1997).
Transient accretion-powered pulsars in outburst show a positive dependence of
1.2 Pulsating neutron stars 9
Table 1.1. Quasi-Periodic Oscillations in Accretion-Powered Pulsarsa
Source Spin Frequency (mHz) QPO frequency (mHz)
4U 1907+09 2.27 55XTE J1858+034 4.5 111A 0535+26 9.71 27–72EXO 2030+375 24 187–213LMC X-4 74 0.65–1.35, 2–204U 1626−67 130 1,48Cen X-3 207 35V 0332+53 229 514U 0115+63 277 2, 62Her X-1 807.9 8, 12, 43SMC X-1 1410 60?
GRO 1744−28b 2140 40000
a compilation after Shirakawa & Lai 2002; b Zhang et al. 1996
the accretion torque (as measured by the spin-up rate) on the inferred accretion
luminosity (Fig. 1.4; left panel). This is consistent with the simple model of disk-magnetosphere interaction (e.g., Ghosh & Lamb 1992), in which, as the accretion
rate increases, the rate of angular momentum transfer from the accretion flow to the
neutron star increases. At the limit of very low mass accretion rate, the neutron stars
are expected to spin down, because the magnetic field lines that couple to the outer,
slower accretion flow remove spin angular momentum from the neutron star. Suchspin-down episodes have not been detected by BATSE, although evidence for spin
down in the pulsar EXO 2030+375 has been previously reported based on EXOSAT
data (Parmar et al. 1989). The very low fluxes down to which the transient sources
continue to spin up place strong constraints on the relative importance of angularmomentum transfer between the accretion disk and the neutron star via anchored
magnetic field lines.
In sharp contrast to the transient sources, persistent disk-fed pulsars show a bi-
modal behavior in their accretion torques (Bildsten et al. 1997; see also Fig. 1.4;right panel). Episodes of spin-up and spin-down of approximately equal accretion
torques alternate at timescales that vary from ∼ 10 days (e.g., in Cen X-3) to ≥ 10 yr
(e.g., in GX 1+4). The transition between spin-up and spin-down is rapid (≤ a few
days) and cannot be resolved with BATSE measurements. Current models of the
disk-magnetosphere interaction in accretion-powered pulsars can account for the ob-served bimodal torques only if one of the physical properties of the accretion flow
is also assumed to show a bimodal behavior. Such assumptions include a bimodal
distribution of the mass transfer rate onto the pulsar, or a bimodal dependence on
accretion rate of the orientation of the disk (Nelson et al. 1998; van Kerkwijk et al.1998), of the orbital angular velocity of the accreting gas (Yi & Wheeler 1998), or of
the strength and orientation of any magnetic field produced in the disk (Torkelsson
1998). Alternatively, for any given mass accretion rate onto the neutron star, two
equilibrium solutions may be possible, one in which the star is spinning up and one in
10 Accreting neutron stars and black holes:a decade of discoveries
Fig. 1.5. The optical (left) and X-ray (right) power spectrum of the accretion-poweredpulsar 4U 1626−67 showing mHz quasi-periodic oscillations (Chakrabarty et al. 2001).
which it is spinning down (Lovelace et al. 1999). It is not clear at this point which, if
any, of these alternatives is responsible for the observed torque reversals in accreting
neutron stars and this remains one of the puzzles of the BATSE monitoring of slow
accretion-powered pulsars.
1.2.1.2 Quasi-periodic oscillations
In several accretion-powered pulsars, the power-density spectra in X-rays or
in longer wavelengths show a number of quasi-periodic oscillations, in addition to the
period of the pulsars (Table 1.1; Fig. 1.5; §2.11). The frequencies of these oscillationsrange from ≃ 1 mHz to ≃ 40 Hz and they can be from ∼ 100 times smaller to ∼ 100
larger than the pulsar spin frequencies.
The frequency of the fast oscillations in the transient pulsar EXO 2030+375 wasfound to be in good agreement with beat-frequency models (Finger et al. 1998),
in which oscillations occur at the beat frequency between the orbital frequency of
matter at the Alfven radius and the stellar spin frequency (Alpar & Shaham 1995).
On the other hand, the low-frequency oscillations observed in 4U 1626−67 appear
also as asymmetric sidebands to the pulse period (Kommers et al. 1998) and areprobably related to a low-frequency modulation of the accretion flow, possibly due
to the presence of a precessing disk warp (Shirakawa & Lai 2002). It is unclear at
this point whether all of these quasi-periodic oscillations are related to the same
phenomenon or not and what is their physical origin.
1.2.1.3 Cyclotron lines
The X-ray spectra of accretion-powered pulsars are typically described in
terms of relatively flat power-laws with exponential cut-offs at energies ≥ 10 keV.
These continuum spectra are believed to be the result of upscattering of soft pho-
1.2 Pulsating neutron stars 11
10-2
10-1
100
101
102
103
coun
ts/s
-keV
10-6
10-5
10-4
10-3
10-2
10-1
phot
ons/
cm2 -s
-keV
4U 0115+63
phase D
11.8
24.134.5
47.0
66.5
-4-2024
Sig
ma
3 10 30 60 100Energy (keV)
Fig. 1.6. The observed (histogram) and model spectrum (solid line) of the accretion-poweredpulsar 4U 0115+63 showing evidence for cyclotron lines with as many as four overtones(Heindl et al. 1999).
tons by the hot electrons in the accretion columns above the magnetic polar caps
(Meszaros 1992). For neutron-star magnetic field strengths of ≃ 1012 G, the cy-
clotron energy on the stellar surface is ≃ 11.6 keV and the continuum spectra are
expected to show evidence for harmonically related “cyclotron resonance scatteringfeatures” (or simply cyclotron lines) in the X-rays. Observation of such features
was anticipated from the early days of X-ray astronomy and expected to lead to
direct measurements of the magnetic field strengths of accreting neutron stars (e.g.,
Trumper et al. 1978).The broadband spectral capabilities of RXTE and BeppoSAX made possible the
unequivocal detection of harmonically related features in four accretion-powered pul-
sars, as well as the detection of single features in ten more sources (Fig. 1.6 and
Table 1.2). The widths of the resonance features appear to be correlated to the
energies of the continuum cut-offs and to be proportional to their central energiesand to the inferred scattering depths (Coburn et al. 2002). The central energies of
these features provided direct measurements of the surface magnetic fields of accret-
ing pulsars, which can be used in constraining the models of disk-magnetosphere
interaction. Future observations of the pulse-phase dependence of the scattering fea-tures is also expected to provide more detailed constraints on the geometries of the
accretion columns in slow accretion-powered pulsars.
1.2.2 Millisecond accretion-powered pulsarsThe presence of accretion-powered millisecond pulsars in low-mass X-ray
binaries had been predicted 17 years before such sources were eventually discovered.
Millisecond rotation-powered pulsars were thought to acquire their low (≃ 108 G)
magnetic fields and fast spin frequencies while accreting mass at high rates in low-
12 Accreting neutron stars and black holes:a decade of discoveries
Table 1.2. Energies of Cyclotron Lines in Accretion-Powered Pulsarsa
Source Spin Period (s) Energy (keV) Harmonics
4U 0115+63 3.61 12 Yes4U 1907+09 438 18 Yes4U 1538−52 530 20 YesVela X-1 283 25 YesV 0332+53 4.37 27Cep X-4 66.2 28Cen X-3 4.82 28.54U 0352+309 835 29XTE J1946+274 15.8 36MX 0656−072 160.7 364U 1626−67 7.66 37GX 301−2 681 37Her X-1 1.24 41A 0535+26 105 50 or 110
a after Heindl et al. 2004
mass X-ray binaries (Alpar et al. 1982; Radhakrishnan & Shrinivasan 1982). These
millisecond radio pulsars were most often found in binaries with evolved, low-mass
white dwarf companions (Bhattacharya & van den Heuvel 1991), which were thoughtto be the descendents of LMXBs, and their birthrates were similar (albeit with
systematic differences) to those of LMXBs (see, e.g., Kulkarni & Narayan 1988;
Lorimer 1995 and references therein). Moreover, circumstantial evidence based on
the bursting behavior (Lewin, van Paradijs, & Taam 1996), rapid variability (Alpar
& Shaham 1985; Ghosh & Lamb 1991), and X-ray spectra (Psaltis & Lamb 1998) ofthe compact objects in bright LMXBs strongly supported their identification with
weakly-magnetic, rapidly spinning neutron stars. However, despite intense searches
(e.g., Vaughan et al. 1994), periodic pulsations could not be detected from any
LMXB, making this the holy grail of X-ray binary astrophysics in the pre-RXTEera.
The discovery, with RXTE, of highly coherent pulsations in the X-ray fluxes of
LMXBs during thermonuclear X-ray bursts (Strohmayer et al. 1996; see also Chap-ter 3 in this volume) provided the then strongest evidence for the presence of neu-
tron stars with millisecond spin periods in LMXBs. However, the first bona fide
millisecond, accretion powered pulsar was discovered only in 1998, in a transient
ultracompact binary (see Fig. 1.7). Since then, four additional millisecond pulsarswere discovered in very similar transient binaries (see Table 1.3).
The spin periods and inferred magnetic fields of these five pulsars are indeed
consistent with the prediction that such systems are the progenitors of the rotation-powered millisecond pulsars observed in radio wavelengths (see, e.g., Fig. 1.8; Psaltis
& Chakrabarty 1998). However, the binary systems which all five pulsars belong to
have a number of unusual characteristics. First, they are all ultracompact as inferred
from their small orbital periods (≤ 2 hr) and projected semi-major axes (≤ 62 lt-
1.2 Pulsating neutron stars 13
Fig. 1.7. The power-density spectrum of the X-ray flux from SAX J1808.4−3658 showingclearly the ∼ 401 Hz pulsation (left; Wijnands & van der Klis 1998) and the pulse-timeresiduals caused by its 2 hr orbit (right; Chakrabarty & Morgan 1998).
Table 1.3. Observed Properties of Millisecond Accreting Pulsarsa
Source fs (Hz) Porb (m) a (lt-ms) f (M⊙)
SAX J1808.4−3658b 401.0 120.9 62.809 3.78 × 10−5
XTE J0929−3314c 185.1 43.6 6.290 2.7 × 10−7
XTE J1751−305d 435.3 42.4 10.1134 1.278 × 10−6
XTE J1807−294e 190.6 40.1 4.80 1.54 × 10−7
XTE J1814−338f 314.3 256.5 390.3 2.016 × 10−3
a Spin frequency, orbital period, projected semi-major axis, and mass function;b Wijnands & van der Klis 1998 and Chakrabarty & Morgan 1998; c Galloway et al.
2002; d Markwardt et al. 2002; e Markwardt et al. 2003; Markwardt, priv. comm.;f Markwardt & Swank 2003; Markwardt, priv. comm.
ms). Moreover, the companions to the neutron stars have masses of only a few
hundredths of the solar mass and are, most probably, the remnants of white dwarfs
that have lost most of their mass. Finally, the inferred long-term averages of the
mass accretion rates onto these neutron stars, as well as their maximum luminositiesduring outbursts are among the lowest in the known LMXB population (see, e.g.,
Chakrabarty & Morgan 1998).
Although they resolved one of the long-standing puzzles in X-ray binary astro-
physics, these five millisecond pulsars have posed two very important questions:
First, why do these particular neutron stars in these special binaries appear as X-
ray pulsars, whereas the compact objects in the other LMXBs do not; and second,why does the spin-up of these systems stop short of the sub-millisecond periods that
most neutron-star models allow (see, e.g., Glendenning 2003 and references therein).
No satisfactory answer to the first question has been found to date, whereas three
equally exciting answers to the second question appear plausible.
Whether an accreting neutron star appears as an X-ray pulsar depends mostly
14 Accreting neutron stars and black holes:a decade of discoveries
Fig. 1.8. Inferred strengths of the dipole magnetic fields of slow radio pulsars (dots),millisecond radio pulsars (open circles), and of the accretion-powered millisecond pulsarSAX J1808.4−3658 (Psaltis & Chakrabarty 1999).
on three factors: its magnetic field, the mass accretion rate, and the relative orien-
tation of the binary, spin, and magnetic axes to the direction of the observer. For
example, unless the binary axis makes a relatively small angle with the directionof the observer, the accretion flow may block the direct viewing of the polar caps
and hence the modulation of the X-ray flux at the stellar spin frequency. Motivated
by the very small measured values of the mass function of SAX J1808.4−3658 (see,
Table 1.3), which suggest a priori small inclination angles, Psaltis & Chakrabarty
(1998) suggested that a favorable viewing geometry is responsible for the fact thatthis source appears as an X-ray pulsar. However, the subsequent discovery of four
additional pulsars in similar binary systems (albeit with equally small mass func-
tions), the discovery of a small modulation of the X-ray flux at the orbital period of
SAX J1808.4−3658 (Chakrabarty & Morgan 1998), as well as the optical propertiesof the reprocessed radiation from the accretion disk (Wang et al. 2000) all argue
against such a favorable geometry.
Alternatively, if the Alfven radius is smaller than the size of the neutron star thenthe stellar magnetic field is nowhere dynamically important and cannot channel the
accretion flow preferentially onto the magnetic poles. Cumming et al. (2001) argued
that in the accreting pulsars the very low rate of mass transfer onto the neutron
stars is not sufficient to “bury” their magnetic fields and make them dynamicallyunimportant, in contrast to the case of the other LMXBs, that are accreting at
much larger rates on average. This is consistent with the peculiar properties of the
binaries in which all five millisecond accreting pulsars reside. It is, however, hard to
reconcile with the fact that these five sources, besides being pulsars, have very similar
1.3 Non-pulsing neutron stars and black holes 15
X-ray spectra (Gilfanov et al. 1998) and aperiodic variability properties (Wijnands
& van der Klis 1998) with many non-pulsing sources.
It appears that, in the five millisecond pulsars, the stellar magnetic fields only
introduce a modulation of the X-ray flux at the stellar spin frequency without altering
significantly the other timing and spectral properties of the systems, and hence their
accretion flows. Consistent with this fact, these five pulsars have spin frequenciesin the same range as the ones inferred in most other, non-pulsing LMXBs from the
properties of burst oscillations (Chakrabarty et al. 2003). Indeed, the spin frequencies
of all these sources are limited to be less than ≃ 700 Hz, which is also similar to
the spin frequency of the fastest known millisecond, rotation-powered pulsar. Thephysics that sets this limit is unclear at this point.
As the simplest explanation, general relativity and the equation of state of neutron-
star matter may not permit stable neutron stars with spin frequencies faster than
≃ 700 Hz. Albeit reasonable, this limit on the spin frequency would require neutronstars to be about twice as large as predicted by most models and hence, if correct,
would point to a more exotic equation of state than any currently discussed (Cook,
Shapiro, & Teukolsky 1994). Alternatively, if the neutron stars in LMXBs spin near
the magnetic spin equilibrium, there may be a natural upper limit on spin frequency
that magnetic accretion can achieve. This would require a lower bound on the stellarmagnetic field, since a non-magnetic neutron star will always spin-up to the maximum
frequency allowed by its equation of state, and hence would provide significant clues
for the efficiency of magnetic field decay in accreting neutron stars.
The most exciting alternative, however, is the possibility that neutron stars spin-ning at rates faster than about 700 Hz rapidly loose their spin angular momenta via
emission of gravitational radiation (Bildsten 1998). For this to happen, the distribu-
tion of matter inside the neutron star must be non-axisymmetric since the emission
of gravitational radiation depends on the time derivative of the mass quadrupole ofthe star. Temperature anisotropies in the surface layers of accreting neutron stars
may result in anisotropies in the crystallization of the material underneath that are
sufficient to account for the rapid loss of spin angular momentum (Bildsten 1998).
Alternatively, excitation of non-radial modes in the neutron star may provide such
a time-dependent mass quadrapole (Andersson, Kokkotas, & Schutz 1998). If this isthe reason why the spin-up of accreting neutron stars stalls at a frequency well below
the maximum frequency allowed by their equation of state, then LMXBs may become
the first sources detected in the very near future by gravitational wave observatories.
1.3 Non-pulsing neutron stars and black holes
The large majority of accreting compact objects show no evidence for peri-
odic pulsations in their persistent emission. In the case of accreting black holes, this
is a direct consequence of the presence of the event horizon, which does not allow forany stable feature to be anchored to the rotation of the compact object. In the case
of neutron stars, however, the absence of pulsations requires a rather weak magnetic
field (≤ 108 G) so that the accretion flow is not disrupted and channeled onto the
magnetic poles.
16 Accreting neutron stars and black holes:a decade of discoveries
Fig. 1.9. Lightcurves of a neutron-star (Aql X-1) and a black-hole (GRO J1655−40) tran-sient source, as observed by the All Sky Monitor on RXTE.
1.3.1 Transient and Persistent sources
Non-pulsing accreting compact objects appear both as persistent and as
transient sources. Members of the first class are observed at X-ray fluxes that can
be variable by up to several factors of two over timescales ranging from millisecondsto months. Their distinguishing characteristic, however, is the fact that they have
been at detectable flux levels for most of the history of X-ray astronomy.
Transient sources, on the other hand, are characterized by long periods of inactiv-
ity, lasting months to decades, that are interrupted by short outbursts, during whichtheir X-ray brightness increases by several orders of magnitude (Bradt et al. 2000).
Figure (1.9) shows typical lightcurves of a neutron-star (Aql X-1) and a black-hole
transient (GRO J1655−40) as observed by the All Sky Monitor on RXTE (cf. §4).
The nature of the compact object in a transient system appears to affect its prop-erties in four ways: the fraction of transients among the black-hole systems is larger
than the fraction of transients among neutron-star systems and their outbursts are
typically longer and rarer (see, e.g., Figure 1.9); moreover, black-hole transients in
quiescence are significantly fainter than their neutron-star counterparts. Accordingto our current understanding, the above differences are caused by the different mass
ratios of the members of the binary systems between the two populations as well as
by the presence of an event horizon in the black-hole systems.
The prevailing model of transient sources is based on the disk instability modelof illuminated accretion disks (van Paradijs 1996; King, Kolb, & Burderi 1996):
accretion flows that extend to large radii (typically > 109 − 1010 cm) from the
compact object have characteristic temperatures less than ∼ 104 K, at which the
anomalous opacity related to the ionization of hydrogen renders them susceptible to
1.3 Non-pulsing neutron stars and black holes 17
Fig. 1.10. The quiescent X-ray luminosity of neutron-star (open circles) and black-holetransients (filled circles) as a function of their orbital period (after Garcia et al. 2001).
a thermal disk instability (see King, §13). At the off-cycle of the instability, material
piles up at the outer edges of the accretion disk with very little, if any, mass accreted
by the central object; this is the quiescent phase of the transient. When the diskbecomes unstable, the accretion flow evolves towards the central object at the viscous
timescale, and the system becomes a bright X-ray source in outburst.
The disk-instability mechanism depends crucially on the temperature of the accre-
tion disk, and hence heating of the disk by illumination can alter the above picture.
The details of illumination by the central object or the disk itself are hard to com-
pute, as they depend strongly on how concave the accretion disk is, on the presenceof warps that can expose different parts of the disk to radiation, and on the effects
of disk winds that can backscatter radiation towards the disk plane (see, e.g., Dubus
et al. 1999). It is generally expected, though, that neutron stars illuminate their
accretion disks more efficiently than black holes , since most of the accretion en-
ergy has to be released near their surfaces. The stabilizing effect of illumination canthen account for the difference in the outburst properties between neutron-star and
black-hole systems.
Studies of the quiescent emission of X-ray transients became possible only after
the launch of X-ray telescopes with good resolution and low background, such as
ASCA and the Chandra X-ray Observatory (see, e.g., Rutledge et al. 2001). It is
now well established that the quiescent emission of black-hole transients is fainter bymore than an order of magnitude compared to the quiescent emission of neutron-star
sources in similar binary systems (Fig. 1.10). The time-variability and non-thermal
character of the X-ray brightness in both cases strongly suggests that at least a
large fraction of the emission arises from accretion (see, e.g., Narayan, Garcia &
18 Accreting neutron stars and black holes:a decade of discoveries
Table 1.4. Masses of Compact Objectsa
Source Period Mass(d) (M⊙)
Black Hole Candidates
GRO J0422+32 0.212 3.66–4.97A0620−00 0.323 8.70–12.86GRS 1009−45 0.285 3.64–4.74XTE J1118+480 0.170 6.48–7.19GS 1124−683 0.433 6.47–8.184U 1535−47 1.116 8.45–10.39XTE J1550−564 1.543 8.36–10.76GRO J1655−40 2.622 6.03–6.57H1705−250 0.521 5.64–8.30SAX J1819.3−2525 2.817 6.82–7.42XTE J1859+226 0.382 7.6–12.0 (?)GRS 1915+105 34 10.0–18.0 (?)GS 2000+25 0.344 7.15–7.78GS 2023+338 6.471 10.06–13.88LMC X-3 1.705 5.94–9.17LMC X-1 4.229 4.0–10.0 (?)Cyg X-1 5.600 6.85–13.25
Accreting Neutron Stars
Vela X-1 8.964 1.78±0.154U 1538−52 3.728 1.06+0.41
−0.34
SMC X-1 3.892 1.17+0.36−0.32
LMC X-4 1.408 1.47+0.44−0.39
Cen X-3 2.087 1.09+0.57−0.52
Her X-1 1.700 1.04+0.75−0.58 or 1.47+0.23
−0.37
Cyg X-2 9.844 1.78±0.23
aBlack hole masses are from Orosz (2002); accreting neutron-star masses are from van
Kerkwijk et al. (1995), Barziv et al. (2001), and Orosz & Kuulkers (1999); radio pulsar
masses are from Thorsett & Chakrabarty (1998).
McClintock 2001). However, release of heat buried in the deep layers of a neutron
star (Brown, Bildsten, & Rutledge 1998) and coronal emission from the companion
star (Bildsten & Rutledge 2000; Campana & Stella 2000; but see Lasota 2000) are also
expected to contribute to the total X-ray brightness. In most cases, the existenceof an event horizon in the black-hole sources, which traps a large fraction of the
accretion luminosity and does not allow the storage of latent heat, is believed to be
responsible for their significantly less luminous quiescent emission (Narayan et al.
2001).
The transient nature of these systems had hampered their systematic study un-
til recently. Since the mid-1990’s, however, the Wide-Field Cameras on BeppoSAX
and the All Sky Monitor on RXTE have revealed and monitored a large number of
1.3 Non-pulsing neutron stars and black holes 19
Table 1.5. Masses of Compact Objects (continued)
Source Period Mass(d) (M⊙)
Double Neutron Stars
J1518+4904 8.634 1.56+0.13−0.44
1.05+0.45−0.11
B1534+12 0.421 1.339±0.0031.339±0.003
B1913+16 0.323 1.4411±0.000351.3874±0.00035
B2127+11C 0.335 1.349±0.0401.363±0.040
B2303+46 12.34 1.30+0.13−0.46
1.34+0.47−0.13
Radio Pulsars in Binaries
J0437−4715 5.741 <1.51J1012+537 0.605 1.7±0.5J1045−4509 4.084 <1.48J1713+0747 67.825 1.45±0.31B1802−07 2.617 1.26+0.08
−0.17
J1804−2718 11.129 <1.73B1855+09 12.327 1.41±0.20J2019+2425 76.512 <1.68J0045−7319 51.169 1.58±0.34
transient sources. The impact of such monitoring programs has been enormous. Thestatistics of the transient sources and their recurrence times shed light on the total
number of X-ray binaries in the Galaxy, their birthrates, and formation mechanisms.
The rise and decay times of their outbursts helped constrain the efficiency of angular
momentum transport in accretion disks (e.g., Hameury et al. 1998). More signif-
icantly, however, optical observations of X-ray transients in quiescence allowed forthe measurement of the masses of many compact objects and have thus provided the
best evidence for the existence of stellar-mass black holes in the Galaxy (McClintock
& Remillard, this volume).
1.3.2 Long-wavelength counterparts
X-ray binaries are initially identified from their intense X-ray brightness and
hard spectra. However, the properties of the binary systems, such as their orbital
periods, the masses of the companion stars, and the masses of the compact objectscan be determined only if a counterpart of the X-ray source in other wavelengths is
also observed (see Charles, §5).
The brightness of luminous X-ray sources in optical wavelengths, especially for
compact objects with low-mass companions, is typically positively correlated with
20 Accreting neutron stars and black holes:a decade of discoveries
Fig. 1.11. Masses of black-hole candidates in binary systems and their orbital periods; thehorizontal dashed line corresponds to the 3.2M⊙ separatrix (see §1.3.8) between neutronstars and black-hole candidates (after Orosz 2002).
their X-ray brightness and orbital periods (van Paradijs & McClintock 1995). This
is a strong indication that thermal emission from the outer parts of the accretion
flow as well as reprocessing of the X-ray emission at the outer accretion disk and the
companion star are responsible for at least part of the optical emission.
Detailed studies of X-ray binaries in IR/optical/UV wavelengths are crucial in
measuring the temperature profiles, ionization fractions, and abundances of elements
in the accretion disk (see, e.g., Kallman et al. 1998) or even the inclinations of the
binary systems (see, e.g., de Jong et al. 1996; Wang et al. 2001). Moreover, if theaccretion geometry is that of a geometrically thin disk, such studies most probably
provide the only handle in measuring the rate at which matter is transfered towards
the compact object (see, e.g., Vrtilek et al. 1990).
In transient systems during their quiescent periods as well as in systems with short
orbital periods, the optical emission is dominated by the companion star, making pos-
sible the measurement of their binary orbits and of the masses of the compact objects.
Table 1.4 and Fig. 1.11 summarize the current mass measurements of compact ob-jects in accreting binary systems and compare them to the mass measurements of
radio pulsars in non-interacting binaries. There is a clear dichotomy between com-
pact objects clustered around ∼ 1.5M⊙ and objects with significantly larger masses.
This is believed to be a direct consequence of the fact that there is an upper limit on
1.3 Non-pulsing neutron stars and black holes 21
Fig. 1.12. Hα spectra (left panels) and reconstructed Doppler images (right panels) of theblack-hole candidates GS 2000+25 (top) and Nova Oph 1977 (bottom; Harlaftis 2001).
the mass of a stable neutron star and any heavier compact object must be a black
hole (but see also §1.3.8).
Recent developments in observational methods and analysis techniques of long-
wavelength data have also led to a number of ways of imaging indirectly the ac-cretion flows around compact objects (see, e.g., Harlaftis 2001). Eclipse mapping
is applicable only to high-inclination systems and uses the periodic occultation of
different parts of the accretion flow by the companion star during the orbit to infer
the relative contribution of different regions to the total emission (see, e.g., Vriel-
mann 2000 for a review of the method and applications to cataclysmic variables).Doppler tomography uses the dependence on orbital phase of Doppler-shifted atomic
lines that originate from different parts of the accretion flow in order to produce a
map of the line-emitting regions on the orbital plane of the binary system (see, e.g.,
Marsh 2000; Harlaftis 2001). Finally, echo tomography uses the time delays betweenprompt and reprocessed emission at different regions in order to map the geometry
of the accretion flows and the binary orbits (see, e.g., Horne 2003).
The above indirect imaging techniques provide the most concrete method of obser-
vationally testing accretion flow models. They have been regularly used in mapping
the radial temperatures of geometrically thin accretion disks, often showing signifi-
cantly flatter profiles compared to accretion theory (see Vrielmann 2000). They can
22 Accreting neutron stars and black holes:a decade of discoveries
MAR 27
APR 03
APR 09
APR 16
APR 23
APR 30
NE
1"
Fig. 1.13. A series of 3.5cm VLA images of the black-hole candidate source GRS 1915+105,showing the fast-traveling knots of the radio jet (Mirabel & Rodriguez 1994).
be used to measure the geometrical thickness of the accretion flows, thereby distin-guishing between competing models (see, e.g., O’Brien et al. 2002). Finally, they
provide the best evidence for the presence of spiral structures in accretion disks,
thus giving clues to the viscous mechanisms that operate in these flows (see, e.g.,
Harlaftis 2001).
1.3.3 Jets
Outflows and collimated jets are ubiquitous phenomena in all accreting ob-
jects, from young stars to active galactic nuclei. The discovery of mildly relativistic
jets from the binary system SS 433 (Spencer 1979) and of superluminal jets fromthe black-hole candidate GRS 1915+105 (Mirabel & Rodriguez 1994; also Fig. 1.13)
revealed that accreting galactic compact objects are no exception to this rule (see
Fender, §9).
Jets from accreting stellar-mass black holes in the galaxy share many properties
with their counterparts in active galactic nuclei. For example, they have non-thermal,
1.3 Non-pulsing neutron stars and black holes 23
polarized radio spectra indicating the presence of shock-accelerated relativistic elec-
trons that emit synchrotron radiation as they propagate in regions with large-scalemagnetic fields (e.g., Mirabel et al. 1998; Eikenberry et al. 1998). They also show
large flux ratios between the approaching and receding sides of the jets, as expected
for relativistic flows (Mirabel & Rodriguez 1994).
In addition to revealing such similarities, however, the short dynamical timescalesassociated with galactic black-hole jets and their proximity made possible an in depth
study of their properties, even though only a handful of jet sources is currently known.
The orientation of some jets in the sky, with the jet in SS 433 being the main example
(Margon 1984), has been observed to precess in real time over large angles with long
periods (∼ 162 days in the case of SS 433). This is believed to be associated tothe precession of the underlying accretion disk (see, e.g., Begelman & Rees 1984;
Ogilvie & Dubus 2001) and may provide clues towards understanding the launching
and collimation of the jets.
The propagation speeds of the jets and their deceleration upon interactions withthe interstellar medium can be measured on images by following the kinematics of
individual radio knots (see, e.g, Fig. 1.13). Not surprisingly, the inferred speeds
appear to depend on the state of the underlying accretion disk from which they are
launched, with steady and transient sources showing mildly and highly relativistic
jets, respectively (see also Fender, §9).Coordinated observations of accreting black holes and neutron stars in the radio,
infrared, and X-rays recently revealed the most striking of jet properties: there is a
very clear correlation between the presence of jets and the X-ray spectral state of
the accretion flows (see, e.g., Corbel et al. 2000). In particular, jets appear when theX-ray spectra of the sources indicate emission from hot electrons (∼ 100 keV; see
next section); on the other hand, when their spectra are typical of cold, geometrically
thin accretion disks, the jets are weak or absent. It is unclear at this point what is
the causal connection between the radio and X-ray properties of accreting compact
objects. The mechanism responsible for the heating of electrons in the accretion flowmay be related to the formation of an outflow, as is the case both for magnetically
active accretion disks (see, e.g., Blandford & Payne 1982) or for advection dominated
accretion flows (Narayan & Yi 1994). Alternatively, most of the X-ray emission may
be produced directly at the base of the jet (Markoff et al. 2001).Finally, recent observations of the X-ray emission from the jet of SS 433 with
Chandra confirmed the presence of atomic emission lines, strongly suggesting that
heavy ions are also accelerated together with the electrons and positrons that are
responsible for most of the jet emission (Marshall et al. 2002). The jet in SS 433,
however, is only mildly relativistic and it is not clear whether this is related to thepresence of ions in the jet or if there simply is an observational selection effect against
detecting lines from very relativistic outflows (Mirabel et al. 1997).
1.3.4 X-ray and γ-ray spectroscopyThe design of high-energy missions with broad-band spectral coverage, the
numerous campaigns of simultaneous observations with multiple instruments, as well
as the advent of high-resolution CCD and grating spectrographs for X-rays have
launched a new era of high-energy spectroscopy in astrophysics. Accreting compact
24 Accreting neutron stars and black holes:a decade of discoveries
0.001 0.010 0.100 1.000Energy (MeV)
0.0001
0.0010
0.0100
0.1000
E F
E
(M
eV
2 c
m-2 s
-1 M
eV
-1)
TTM+HEXE
GRO J0422+32
ASCA
GRS 1716-249
ASCA
GRO J1655-40
Fig. 1.14. Broad-band spectra of black-hole candidates: (left) The infrared to soft X-rayspectrum of XTE J1118+480 (McClintock et al. 2001); (right) the soft X-ray to γ-rayspectra of three sources (Grove et al. 1998).
objects are often being monitored in all wavelengths, providing a strong handle on
their bolometric luminosities as well as placing stringent constraints on accretionmodels.
Figure 1.14 shows some examples of broad-band spectra of black-hole candidates,
from the infrared to the γ-rays. All such spectra of accreting compact objects showunequivocally that a number of distinct emission mechanisms are responsible for
their various features. In fact, simple thermal emission models from geometrically
thin accretion disks can produce neither the spectral complexity nor the hard X-
ray and γ-ray fluxes that are observed. Modeling the spectra of accreting compactobjects is complicated and appears to be mostly data driven. However, it is also
potentially very rewarding as it may lead to the understanding of processes such
as the generation of magnetic fields in turbulent flows, viscous heating in magnetic
media (see, e.g., Quataert & Gruzinov 1999), and the thermal properties of multi-
temperature plasmas (e.g., Coppi 1999).
Most current models of the X-ray emission from accreting neutron stars and black
holes (see GRO J0422+32 and GRS 1716−249 in Fig. 1.14) require that a tenu-
ous atmosphere of hot (∼ 10 − 100 keV) electrons is present simultaneously withthe geometrically thin accretion disks. Such a hot medium may be in the form
of a magnetically heated corona (see, e.g., Dove, Wilms, & Begelman 1997), of an
advection-dominated accretion flow (ADAFs; Esin, McClintock, & Narayan 1997),
or even a jet (Markoff, Falcke & Fender 2001). These same electrons are almost cer-tainly responsible also for the longer-wavelength emission (see XTE J1118+480 in
Fig. 1.14), which appears to be strongly correlated with the X-ray flux, via radiation
processes associated with the magnetic fields generated and sustained by the accre-
tion flow. Finally, the power-law γ-ray spectral tails of luminous neutron-star and
1.3 Non-pulsing neutron stars and black holes 25
Fig. 1.15. High resolution X-ray spectra of (top) the ADC source 4U 1822−37 (Cottam etal. 2001) and (bottom) the black-hole source Cyg X-1 as observed by the Chandra X-rayObservatory (Miller et al. 2002).
black-hole candidates (see GRO J1655−40 in Fig. 1.14; see also Grove et al. 1998; Di
Salvo et al. 1999) require the existence of a non-thermal population of very energetic
electrons, either in the form of a hybrid plasma (Coppi 1999) or of a quasi-radialhigh-velocity flow (Laurent & Titarchuk 1999).
Perhaps the most eagerly anticipated result of the launch of X-ray telescopes withhigh spectra resolution, such as ASCA, Chandra, and XMM/Newton, has been the
discovery of atomic lines from the spectra of accreting compact objects. The rela-
tive strengths and equivalent widths of such lines depend strongly on temperature,
density, and ionization flux and hence are valuable probes of the physical conditionsin the accretion flows (Liedahl 1999). Moreover, the gravitational redshifts and rel-
ativistic broadening of atomic lines generated close to the event horizons of black
holes can, in principle, be used to map the spacetimes around the compact objects
and measure properties such as their masses and spin angular momenta (e.g., Fabian
et al. 1989).
The X-ray spectra of many accreting compact objects have, unfortunately, atomic
26 Accreting neutron stars and black holes:a decade of discoveries
lines that are very weak or even undetectable, largely due to their high temperatures
and photonization fluxes. However, studies of atomic lines have been proven fruitfulin several cases for which the binary configurations are optimal. For example, in
accretion-disk corona (ADC) sources, the high inclinations of the binary systems
allow for a clear view of the coronal structure away from the central object, where
the temperatures are lower and the line emission stronger (see Fig. 1.15; also Cottamet al. 2001). In binary systems with companions that exhibit strong winds, the X-ray
spectra show a variety of absorption lines and edges that originate at the relatively
cooler wind material (Fig. 1.15 and, e.g., Miller et al. 2002a). Recent observations of
several black-hole candidates have also shown evidence for relativistically broadened
iron K lines, similar to those observed in active galactic nuclei (Miller et al. 2002b).Finally, a detection of gravitationally redshifted atomic lines during thermonuclear
flashes on the surface of an accreting neutron star has also been recently reported
(Cottam et al. 2002).
1.3.5 Variability
Accreting compact objects are among the most variable persistent sources
in the sky. Even excluding the X-ray transients, the flux from most sources typically
varies by factors of two over periods ranging from months to fractions of a second (see
Fig. 1.16). This is not surprising, given the wide range of characteristic timescalesthat are involved in the processes that lead to the production of the high-energy
emission. For example, the transfer of mass from the binary companion to the
compact object is expected to vary at timescales comparable to the orbital period,
i.e., hours to days. The inward diffusion of matter in the accretion flow occurs atthe viscous timescale, which is slower than the Keplerian orbital frequency at any
radius and ranges from days, at the outer edge of the disk, to fractions of a second,
close to the compact object. Finally, the interaction of the accretion flow with the
central star occurs at the fastest dynamical timescale in the accretion flow, which is
of order of a millisecond.Over the last two decades, the most unexpected result of timing studies of ac-
creting compact objects has been the discovery of quasi-periodic oscillations (QPO;
see Fig. 1.17) of their X-ray brightness at all these timescales (van der Klis 2000,
and §2). Because of the high degree of variability of the sources as well as due toobservational constraints, the fast (greater than a fraction of a Hz) oscillations of the
X-ray brightness are the ones that have been studied more extensively, mostly with
the proportional counters onboard EXOSAT, GINGA, and RXTE.
The large majority of these oscillations have frequencies that are highly variable,
even though they may loose coherence only after tens or hundreds of cycles. Whenseveral variable-frequency QPOs are observed simultaneously, their frequencies follow
a small number of tight correlations (Psaltis, Belloni, & van der Klis 1998). On
the other hand, in several luminous black-hole candidates, pairs of high-frequency
QPOs have also been detected, with frequencies that are nearly constant and in smallinteger ratios (i.e., 3:2, 5:3; Strohmayer 2001a, 2001b; Abramowicz & Kluzniak 2001).
These distinct and correlated frequencies of the observed QPOs clearly suggest that
accretion flows are capable of picking, out of a large pool of alternatives, only a small
number of characteristic frequencies at which to vary preferentially.
1.3 Non-pulsing neutron stars and black holes 27
Fig. 1.16. The light curve of the neutron-star source Cyg X-2, as observed at differenttimescales by the All Sky Monitor (upper two panels) and the Proportional Counter Array(bottom panel) onboard RXTE .
A number of theoretical interpretations have been put forward so far, in an at-
tempt to account for the mechanism that picks these characteristic frequencies in theaccretion flows. Current models of the variable-frequency QPOs require the presence
of a characteristic radius, across which the properties of the accretion disks change
considerably. The observed QPO frequencies are then attributed to the characteris-
tic dynamic or hydrodynamic frequencies of the accretion flow at that radius (e.g.,Stella, Vietri & Morsink 1999; Titarchuk et al. 1999; Psaltis & Norman 1999) or,
in the case of neutron stars, to the coupling of these frequencies to the stellar spin
(Miller, Lamb & Psaltis 1998). On the other hand, models of the constant-frequency
QPOs in black-hole systems are based on the trapping of oscillatory modes in ac-
cretion disks caused by the properties of the relativistic spacetime (Wagoner 1999;Kato 2001) or on the non-linear coupling of different frequencies near the black-hole
event horizons (Abramowicz & Kluzniak 2001).
A common feature of all current models of QPOs is the identification of at least
some of the observed frequencies with dynamic frequencies in the accretion flows, such
as the ones related to the azimuthal orbital motion of plasma, to Lense-Thirring pre-cession, etc. These frequencies correspond to regions very close to the neutron-star
surface or the black-hole event horizons and hence offer the possibility of observing,
for the first time, effects that occur only in strong gravitational fields. In particular,
the general relativistic prediction of a radius, inside which no stable circular orbits
28 Accreting neutron stars and black holes:a decade of discoveries
Fig. 1.17. A power-density spectrum of the neutron-star source Sco X-1 (in two commonlyused representations) showing multiple, simultaneously detected quasi-periodic oscillations.
exist (the so-called innermost stable circular orbit or ISCO), is of fundamental im-portance for almost all models. This characteristic radius is responsible, in different
models, for the saturation of the observed QPO frequencies with accretion rate (e.g.,
Miller et al. 1998; Zhang et al. 1998) or the trapping of modes in the inner accretion
flows and hence the generation of the QPOs themselves (Wagoner 1999; Kato 2001).
More importantly, the azimuthal orbital frequency at the radius of the innermost
circular orbit is thought to provide a natural upper limit on the frequency of any
dynamical process that can occur in an accretion flow (see, e.g., Kluzniak, Michelson& Wagoner 1990; Miller et al. 1998). Since this azimuthal frequency depends only on
the mass and spin of the compact object, it provides, when compared to an observed
QPO frequency, an upper limit on the mass (modulo the spin) of the neutron star or
black hole. This is illustrated as a horizontal dashed line in Figure 1.18, for the case
of the neutron-star source 4U 1636−36. An additional upper limit on the radius ofthe neutron-star (as a function of its mass) is imposed by the requirement that the
maximum observed QPO frequency is less than the azimuthal orbital frequency at
the stellar surface (see Fig. 1.18). These two arguments, together with additional
bounds imposed by the presence of oscillations during thermonuclear bursts (Nathet al. 2002), the detection of redshifted lines from the stellar surfaces (Cottam et
al. 2002), and the measurement of the mass of the compact objects in the binaries
using orbital dynamics will be able to provide the most stringent constraints on the
properties of neutron-star matter and its equation of state.
As discussed above, the rapid variability properties of accretion flows provide useful
probes into the physical conditions close to the compact objects. At the same time,
the slow (≥ 1 day) variability of the same systems can be used in constraining models
1.3 Non-pulsing neutron stars and black holes 29
Fig. 1.18. Constraints on the mass and radius of the neutron star in 4U 1636−36 imposed bythe observation of a 1193 Hz QPO (dashed line; Jonker, Mendez & van der Klis 2002; Milleret al. 1998) and the large amplitudes of burst oscillations (dotted line; Nath, Strohmayer& Swank 2002). The hatched region represents the area of the parameter space that isconsistent with all observations and includes stars with baryonic masses larger than ≃1.4M⊙. The solid lines correspond to neutron-star models with different equations of state.Stellar models with condensates correspond to mass-radius relations with a characteristicflattening at small radii; models of strange stars correspond to mass-radius relations thatstart at the origin.
of the accretion flows at large distances and of the mass transfer process between
the members of the binary. Such studies have become possible recently with the
systematic observation of the entire X-ray sky using the All Sky Monitor onboard
RXTE .
The long-term variability of accreting weakly-magnetic neutron stars and black
holes is typically aperiodic and reflects the variable nature of the mass-transfer and
accretion processes. Sources with more systematic slow variability fall in three cat-
egories: (a) systems with large amplitude variations of their X-ray flux because of
their transient nature (see Fig. 1.9), (b) systems with a periodic modulation at theirorbital frequency caused e.g., by eclipses, by the reflection of the X-ray photons off
the binary companion and the accretion stream, or by the variable rate of mass
transfer due to a highly elliptical orbit (e.g., Cir X-1: Shirey et al. 1996), and (c)
systems with quasi-periodic modulations at variable, superorbital periods (see, e.g.,Wijnands, Kuulkers & Smale 1996; Heinz & Nowak 2001).
Figure 1.19 shows an example of a variable period modulation in the source Cyg X-
2. This is a neutron-star system with a binary period of 9.84 days, which shows a
number of superobital periods, some of which are nearly integer multiples of 9.84 days
30 Accreting neutron stars and black holes:a decade of discoveries
Table 1.6. Long-term periods of X-ray binaries
Source Orbital Period Long-term Period References(d) (d)
Stable
LMC X-4 1.4 30.4Her X-1 1.7 35SS 433 — 164a 16
Quasi-periodic
SMC X-1 3.89 50–60 9LMC X-3 1.7 ∼100–300 11, 12Cyg X-1 5.6 294 13Cyg X-2 9.84 ∼70–80 1, 2, 3, 14, 15Cen X-3 2.09 ∼120 74U 1728−34 — ∼30–70 3, 44U 1820−30 0.008 ∼170 6, 104U 1916−053 0.035 ∼80 54U 2127+119 0.71 ∼37 (?) 8
aas inferred from the precession of the jet;
References: 1. Kuulkers et al. 1996; 2. Paul et al. 2000; 3. Kong et al. 1998; 4. Galloway
et al. 2003; 5. Homer et al. 2001; 6. Chou & Grindlay 2001; 7. Priedhorsky & Terrell
1983; 8. Corbet et al. 1997; 9. Wojdowski et al. 1998 10. Priedhorsky & Terrell 1984; 11.
Cowley et al. 1991; 12. Wilms et al. 2001; 13. Priedhorsky et al. 1983; 14. Smale &
Lochner 1992; 15. Kuulkers et al. 1999; 16. Margon & Anderson 1989.
(see also Wijnands et al. 1996). Such modulation may be related to variable mass
transfer of the binary companion that is driven at the binary orbital period or to
reflection off the warped surface of the outer accretion disk. Recent theoreticalinvestigations of the warping of geometrically thin accretion disks caused by the
torque of the reflecting X-ray irradiation from the central object or by an asymmetric
wind (see, e.g., Pringle 1996; Malloney, Begelman & Pringle 1998) have indeed shown
the possibility of long-lived warping modes that could produce, in principle, the
observed modulations.
1.3.6 Thermonuclear Bursts
The material that is accreted on the surface of a weakly-magnetic neutron
star may be compressed to densities and temperatures for which the thermonuclear
burning of helium is unstable. The ignition of helium results in a rapid (∼ 1 s)
increase in the X-ray luminosity of the neutron star, followed by a slower (∼tensof seconds) decay that reflects the cooling of the surface layers that ignited (see,
Fig. 1.20). The observational manifestation of these thermonuclear flashes are called
Type I X-ray bursts (for a review see Lewin et al. 1996 and Strohmayer & Bildsten,
§3).
For very energetic bursts, the force of the escaping radiation balances gravity,
1.3 Non-pulsing neutron stars and black holes 31
Fig. 1.19. Dynamical periodogram of the neutron-star source Cyg X-2, showing the variable,superorbital oscillations of its X-ray flux, as observed by the All Sky Monitor on RXTE(courtesy D. Galloway).
causing the surface layers of the neutron star to expand rapidly. During the expansion
phase, the emerging radiation flux remains comparable to the Eddington criticalvalue, at which radiation and gravitational forces are balanced, and the remaining
energy of the explosion is given as kinetic and potential energy to the expanding
layers (Kato 1983; Nobili et al. 1994). These are the so-called Eddington-limited
or photospheric radius-expansion bursts. Because the maximum observed flux of anEddington-limited burst depends, to zeroth order, on the mass and radius of the star
(the allowed values of which span a very narrow range) as well as on the distance to
the source, these burst can be very useful in constraining all these three parameters
and, in particular, in measuring the distances to X-ray bursters (see Kuulkers et al.
2003; Galloway et al. 2003).
The general properties of Type I X-ray bursts, such as their energetics, peak fluxes
and fluences, recurrence times, rise and decay timescales, etc. are qualitatively consis-
tent with the predictions of the helium ignition model. Quantitatively, however, the
observations show very little of the regularity that is inherent in the numerical models(Lewin et al. 1996; see, however, Galloway et al. 2004). This is not surprising given
the strong dependence of the burst properties on the time-variable physical conditions
in the ignition area. For example, the energetics and timescales of bursts depend
on the composition of the accreting material (see, e.g., Cumming & Bildsten 2001),the local rate of accretion, the temperature of the neutron-star core (e.g., Fushiki
& Lamb 1987), the presence of ashes from previous bursts, etc. Recently, there has
been significant progress towards developing models of thermonuclear flashes that re-
lax many of the limiting assumptions of earlier calculations and incorporate detailed
32 Accreting neutron stars and black holes:a decade of discoveries
Fig. 1.20. A typical Type I X-ray burst from the neutron-star source 4U 1636−36. Theflat top of the burst lightcurve is characteristic of Eddington-limited (or radius-expansion)bursts (courtesy D. Galloway).
nuclear networks (Schatz et al. 2000), the effects of multi-dimensional propagation
of burning fronts (Zingale et al. 2001; see also Fig. 1.21), and the stellar rotation
(Spitkovsky et al. 2002).
The long-term monitoring capabilities of BeppoSAX and RXTE brought the dis-
covery of a new type of long (∼hr) bursts with even longer (∼yr) recurrence times,
the so-called superbursts (Cornelisse et al. 2000; Strohmayer & Bildsten, §3). These
are believed to be caused by unstable carbon burning in layers that are deeper than
those responsible for the normal Type I bursts, thereby accounting for their longerdurations and recurrence timescales (Strohmayer & Brown 2002; Cumming & Bild-
sten 2002).
During many of the normal Type I X-ray bursts and in one superburst, highlycoherent oscillations of the observed X-ray fluxes are often detected (Strohmayer et
al. 1996; Strohmayer and Bildsten, this volume). The frequencies of the oscillations
drift by a few percent during the bursts, reaching values that are constant, to within
one part in 104, between bursts from the same source (Muno et al. 2002a). In burstsfrom two ultracompact millisecond pulsars, in which the spin frequencies of the stars
are known, the asymptotic values of the burst oscillation frequencies are nearly equal
to the spin frequencies of the stars (see, e.g., Chakrabarty et al. 2003 and §2).
These two properties have led to an interpretation of burst oscillations in which
the thermonuclear burning on the neutron star surface is non-uniform and produces
a modulation of the X-ray flux at the stellar spin frequency (Strohmayer et al. 1996).
The origin of the frequency drift during the rising phase of the bursts, however, is
1.3 Non-pulsing neutron stars and black holes 33
Fig. 1.21. Two-dimensional simulation of helium detonation on a neutron star. Thegreyscale levels correspond to different densities (Zingale et al. 2001)
still unresolved. A number of ideas are currently being explored, which involve the
decoupling and slowing down of the surface layers from the rest of the neutron star
during the rise (Cumming et al. 2002), the drift of the burning front with respect toa fiducial azimuth on the stellar surface caused by rotation (Spitkovsky et al. 2001),
and the excitation of non-radial modes in the burning layers (Heyl 2004).
Despite the lack of a physical model of burst oscillations, their properties have al-
ready been used in obtaining stringent constraints on the masses and radii of neutron
stars, as well as on the degree of non-uniformity of the thermonuclear burning. Forexample, during burst rise, the large (∼ 70%) observed amplitudes of the oscillations
require that the neutron stars are not too compact for gravitational self-lensing to
wash out the pulsations (Nath et al. 2002). Moreover, the lack of detectable harmon-
ics in burst oscillations constrains the emission areas and orientations with respectto the rotation axes of the stars (Muno et al. 2002b).
1.3.7 A Census of Non-Pulsing Neutron Stars and Black Holes
As it has become evident from the discussion in the previous section, non-
pulsing neutron stars and black holes are found in a variety of binary systems and
configurations. The number of known sources has steadily increased, from 33 in1983, to 119 in 1995, to 150 in 2000 (Liu et al. 2001). This increase is largely due to
the discovery of low-luminosity sources with detectors of increasing sensitivity, but
also due to the discovery of a large number of transient sources.
Out of 150 sources, 63 exhibit Type I X-ray bursts and can be, therefore, identified
34 Accreting neutron stars and black holes:a decade of discoveries
Fig. 1.22. A Venn diagram of the known non-pulsing neutron stars and black holes in binarysystems (using the catalog of Liu et al. 2001). The initials correspond to B: bursters; G:globular-cluster sources; T: transients; D: dippers. The areas of the circles correspondapproximately to the relative number of sources in each category.
as neutron stars. On the other hand, 17 sources have dynamically measured masses
in excess of 3.2M⊙ and can be safely identified as black holes. Half of the known
sources (76 out of 150) are transients. Given their relatively low duty cycles, whichis of order 10% for neutron-star sources and much lower for black holes, the total
number of transient sources in the galaxy must be significantly larger (King & Kolb
1996).
Finally, about 10% of the objects in each category (5/63 bursters, 4/76 transients,11/150 total) are dippers. This is consistent with the geometric interpretation of
the dips and with an opening angle of the accretion stream (as measured from the
central objects) of about 10 degrees.
1.3.8 The Nature of the Central Object in Non-Pulsing X-ray Binaries
The presence of persistent pulsations or of Type I bursts in the X-ray
lightcurve of an accreting compact object provide the only unequivocal proof that
the central object is a compact star and not a black hole. However, the absence ofpulsations or bursts does not provide proof that the compact object does not have
a surface (see, however, Narayan & Heyl 2002). Indeed, if the magnetic field of the
neutron star is dynamically unimportant (≪ 109 G), it will not appear as an X-ray
pulsar. Moreover, Type I X-ray bursts are expected to occur only for a particularrange of accretion rates, surface gravity accelerations, compositions of the accreting
material, and core temperatures (see, e.g., Fushiki & Lamb 1987). Finally, the pres-
ence of pulsations or bursts indicates simply the existence of a stellar surface and
does not necessarily prove that the object is a neutron star, as opposed to another
1.3 Non-pulsing neutron stars and black holes 35
type of compact star with yet-to-be-discovered properties. It is in fact extremely dif-
ficult to provide a conclusive proof that a non-pulsing object is a black hole or even aneutron star. All current arguments are either empirical or simply attempted proofs
by elimination, leaving open the possibility that a viable alternative was simply not
considered (see also Abramowicz et al. 2002).
Since the discovery of X-ray binaries, a number of empirical tests have been putforward for distinguishing non-pulsing neutron stars from black holes based on their
X-ray properties. Recent examples of suggested evidence for black holes are the
presence of a hard X-ray spectrum at high luminosities (Barret et al. 1996), of a
particular type of X-ray spectrum (Done & Gierlinski 2003), or of significant variabil-
ity power at high Fourier frequencies (Sunyaev & Revnivtsev 2000). Albeit useful asindicators, these empirical test cannot provide conclusive arguments for the nature
of the central objects. Neutron stars and black holes share the general characteris-
tics of a very complex phenomenology of spectral and timing properties and their
differences are only in the details (see, e.g., van der Klis 1994 for a discussion).The measurement of a large gravitational mass for the compact object is currently
considered to be the strongest evidence for its identification with a black hole. The
reason is that, under three simple assumptions, an optimal upper bound on the mass
of any neutron star of ≃ 3.2M⊙ can be derived (Rhodes & Ruffini 1983). The three
assumptions are: (i) the star is non-rotating; (ii) the equation of state at densitiesbelow some fiducial value (typically close to the nuclear saturation density) is known;
(iii) the speed of sound at larger densities is smaller than the speed of light (the so-
called “causality” condition).
Including the effects of rotation introduces only small corrections and affects thelimiting mass by < 20% (Friedman & Ipser 1987). The next two assumptions,
however, are significantly more constraining. The speed of sound, being a phase
velocity, is not bounded by relativity to be less than the speed of light. In fact,
the actual condition used is a causality requirement only for cold, non-dispersive
material. However, neutron-star matter can be both dispersive and of non-zerotemperature. Relaxing this condition and allowing for rotation leads to bounds on
the neutron star mass as large as ∼ 14M⊙ (Sabbadini & Hartle 1977). Finally, the
compact stars under consideration may form a distinct family of objects, which is
not a continuation of the normal white-dwarf to neutron-star sequence of equilibriumconfigurations towards higher central densities. If this is the case, then the second
of the above assumptions is irrelevant.
Families of compact objects that are not bound by gravity, such as strange star
(Witten 1984; Alcock et al. 1986) and Q stars (Bahcall et al. 1990), have been recently
constructed as potential alternatives to neutron stars and black holes. Strange starscan be made practically indistinguishable from neutron stars with respect to masses,
radii, and maximum spin frequencies (see, however, Glendenning 1997). On the
other hand, Q stars can be constructed to have masses as large as the most massive
stellar-mass black-hole candidates, even though this would require extreme changesin our understanding of the properties of matter at densities as low as one tenth of
nuclear saturation (Miller et al. 1998). Finally, if gravity is not described by general
relativity in the strong-field regime, then the limiting mass of a neutron star may
not be ∼ 3.2M⊙. Metric theories of gravity that are consistent with all solar system
36 Accreting neutron stars and black holes:a decade of discoveries
T
µ
early universe
ALICE
<ψψ> > 0
SPS
quark-gluon plasma
hadronic fluid
nuclear mattervacuum
RHICTc ~ 170 MeV
µ ∼ o
<ψψ> > 0
n = 0
<ψψ> ∼ 0
n > 0
922 MeV
phases ?
quark matter
neutron star cores
crossover
CFLB B
superfluid/superconducting
2SC
crossover
Fig. 1.23. The proposed phase diagram of QCD, showing the position on it of matter in theearly universe, in modern heavy-ion colliders (SPS, RHIC, and ALICE), and in the coresof neutron stars (for details see Hands 2001).
tests but deviate from general relativity in the strong-field regime allow for neutronstars with significantly larger mass (see, e.g., DeDeo & Psaltis 2003).
It is important to note here that the existence of black holes is a strong-field pre-
diction of a theory (i.e., general relativity) that has been tested to high accuracy, at
least in the weak-field limit (Will 2000). On the other hand, all the other alternatives
discussed above are the results of theoretical assumptions that have not been tested(and mostly could not have been tested) with current experiments. Such alternatives
provide physically consistent counter-examples to the identification of a compact ob-
ject as a black hole. However, they will remain simply as thought experiments until
experimental evidence shows that our theories of gravity and matter fail to describeextreme physical conditions.
1.4 Accretion-powered X-ray sources in the 21st century
Accreting neutron stars and black holes in the galaxy offer the unique oppor-tunity of understanding the properties of matter, electromagnetic, and gravitational
fields beyond the conditions found in current terrestrial experiments and other cos-
mic settings. Indeed, matter in the cores of neutron stars occupies a place in the
proposed QCD phase diagram that is distinct from the regions occupied by matterin the early universe and in modern heavy-ion colliders (Fig. 1.23; Hands 2001).
At the same time, the gravitational fields probed by the accretion flows just out-
side the event-horizons of black holes and the surfaces of neutron stars, are many
orders of magnitude stronger than those probed by other tests of general relativity
1.4 Accretion-powered X-ray sources in the 21st century 37
Fig. 1.24. The potential of the gravitational fields probed by different astrophysical obser-vations and tests of General Relativity (after Psaltis 2004).
(Fig. 1.24; Psaltis 2004). X-ray astronomy and the discovery of accretion-powered
neutron stars and black holes provide probes with which tests of basic physics the-
ories can be performed in a way that is complementary to other experiments and
cosmological studies.The last four decades have been the period of discovery, in which the astrophysical
properties of accreting compact objects were investigated. For his contribution to
this effort, Riccardo Giaconni was awarded the 2002 Nobel price in physics. In the
near future, the observations of neutron stars and black holes with detectors with
large surface areas, high spectral resolution, and fast timing capabilities will allow forprecise measurements of the physical conditions in the accretion flows, as they vary
at the dynamical timescales near the compact objects. Moreover, the increase in the
computational power and storage capabilities of supercomputers will allow for the
development of new tools for modeling radiation-magneto-hydrodynamic phenomenain curved spacetimes. And, as it has always been the case in compact-object astro-
physics, this interplay between theory and observations will offer us a more complete
picture of our universe.
Acknowledgements: It is my pleasure to thank a number of people that have
helped me in understanding the concepts described in this chapter; I thank especiallyT. Belloni, D. Chakrabarty, S. DeDeo, D. Galloway, E. Kuulkers, F. Lamb, C. Miller,
M. Muno, R. Narayan, F. Ozel, and M. van der Klis. I am also grateful to Deepto
Chakrabarty for help in planning, writing, and proofreading this chapter, to Duncan
Galloway for producing several of the figures, as well as to Martin Pessah and ErikKuulkers for carefully reading the manuscript.
38 Accreting neutron stars and black holes:a decade of discoveries
References
Abramowicz, M.A., and Kluzniak, W. (2001), A&A 374, L19
Abramowicz, M.A., Kluzniak, W., and Lasota, J.-P. (2002), A&A 396, L31Alcock, C., Farhi, E., and Olinto, A. (1986), ApJ 310, 261Alpar, M. A., Cheng, A. F., Ruderman, M. A., and Shaham, J. (1982), Nature 300, 728
Alpar, M. A., and Shaham, J. (1985), Nature 316, 239Andersson, N., Kokkotas, K., and Schutz, B. F. (1999), ApJ 510, 846
Armitage, P. J., Reynolds, C. S., and Chiang, J. (2001), ApJ 548, 868Bahcall, S., Lynn, B.W., and Selipsky, S. B. (1990), ApJ 362, 251
Balbus, S. A., and Hawley, J. F. (1991), ApJ 376, 214———. (1998), Rev. Mod. Phys. 70, 1
Balbus, S. A., and Papaloizou, J. C. B. (1999), ApJ 521, 650Barret, D., McClintock, J. E., and Grindlay, J. E. (1996), ApJ 473, 963
Barziv, O., Kaper, L., van Kerkwijk, M. H., Telting, J. H., and Van Paradijs, J. (2001), A&A 377,925
Begelman, M. C., and Rees, M. J. (1984), MNRAS 206, 209
Bhattacharya, D., and van den Heuvel, E. P. J. (1991), Phys. Rep. 203, 1Bildsten, L. (1998), ApJ 501, L89
Bildsten, L., and Rutledge, R. E. (2000), ApJ 541, 908Bildsten, L. et al. (1997), ApJS 113, 367
Blandford, R. D., and Begelman, M.C. (1999), MNRAS 303, L1Blandford, R. D., and Payne, D. G. (1982), MNRAS 199, 883Bradt, H., Levine, A.M., Remillard, R.A., and Smith, D.A. (2000), in Multifrequency Behaviour
of High Energy Cosmic Sources, ed. F. Giovannelli and L. Sabau-Graziati, astro-ph/0001460
Brown, E. F., Bildsten, L., and Rutledge, R.E. (1998), ApJ 504, L95Campana, S., and Stella, L. (2000), ApJ 541, 849
Chakrabarty, D., Homer, L., Charles, P. A., and O’Donoghue, D. (2001), ApJ 562, 985Chakrabarty, D., and Morgan, E.H. (1998), Nature 394, 346
Chakrabarty, D., et al. (2003), Nature 424, 42Chou, Y., and Grindlay, J. E. (2001), ApJ 563, 934
Coburn, W., Heindl, W. A., Rothschild, R. E., Gruber, D. E., Kreykenbohm, I., Wilms, J.,Kretschmar, P., and Staubert, R. (2002), ApJ 580, 394
Cook, G., Shapiro, S. L., and Teukolsky, S.A. (1994), ApJ 424, 823
Coppi, P. S. (1999), in High Energy Processes in Accreting Black Holes, ed. J. Poutanen and R.Svensson (ASP: San Fransisco), 375
Corbel, S., Fender, R. P., Tzioumis, A. K., Nowak, M., McIntyre, V., Durouchoux, P., and Sood,R. (2000), A&A 359, 251
Corbet, R. (1997), IAUC 6632Cornelisse, R. et al. (2000), A&A 392, 885
Cottam, J., Paerels, F., and Mendez, M. (2002), Nature 420, 51Cottam, J., Sako, M., Kahn, S. M., Paerels, F., and Liedahl, D. A. (2001), ApJ 557, L101
Cowley, A.P. et al. (1991), ApJ 381, 526Cumming, A., and Bildsten, L. (2000), ApJ 559, L127Cumming, A., Zweibel, E., and Bildsten, L. 2001, ApJ 557, 958
Cumming, A., Morsink, S.M., Bildsten, L., Friedman, J.L., and Holtz, D. E. (2002), ApJ 564,343, L127
DeDeo, S., and Psaltis, D. (2003), PRL 90, 1101
de Jong, J. A., van Paradijs, J., and Augusteijn, T. (1996), A&A 314, 484De Villiers, J., Hawley, J. F., and Krolik, J. H. (2003), ApJ 599, 1238
Di Salvo, T. et al. (2000), ApJ 544, L119Done, C., and Gierlinski M. (2003), MNRAS 342, 1041
Dove, J. B., Wilms, J., and Begelman, M.C. (1997), ApJ 487, 747Dubus, G., Lasota, J., Hameury, J., and Charles, P. (1999), MNRAS 303, 139
Eikenberry, S. S., Matthews, K., Murphy, T. W., Nelson, R. W., Morgan, E. H., Remillard, R. A.,and Muno, M. (1998), ApJ 506, L31
1.4 Accretion-powered X-ray sources in the 21st century 39
Esin, A.A., McClintock, J. E., and Narayan, R. (1997), ApJ 489, 865Fabian, A. C., Rees, M. J., Stella, L., and White, N.E. (1989), MNRAS 238, 729
Finger, M. H., Wilson, R. B., and Harmon, B. A. (1996), ApJ 459, 288Friedman, J. L., and Ipser, J.R. (1987), ApJ 314, 594
Frank, J., King, A.R., and Raine, D. J. (2002), Accretion Power in Astrophysics, (CambridgeUniversity Press; 3rd edition)
Fushiki, I., and Lamb, D. Q. (1987), ApJ 323, L55
Galloway, D., Psaltis, D., Chakrabarty, D., and Muno, M. (2003), ApJ 590, 999Galloway, D. K., Cumming, A., Kuulkers, E., Bildsten, L., Chakrabarty, D., and Rothschild, R. E.
(2004), ApJ 601, 466Galloway, D., Chakrabarty, D., Morgan, E. H., and Remillard, R.A. (2002), ApJ 576, L137
Gammie, C. F., McKinney, J. C., and Toth, G. (2003), ApJ 589, 444Garcia, M.R., McClintock, J. E., Narayan, R., Callanan, P., Barret, D., and Murray, S. S. (2001),
ApJ 553, L47
Ghosh, P., and Lamb, F. K. (1991), ApJ 234, 296———. (1991), in Neutron Stars: Theory and Observations, ed. J. Ventura and D. Pines
(Kluwer), 363
———. (1992), in X-ray Binaries and Recycled Pulsars, ed. E.P. J. van den Heuvel and S. A.Rappaport (Dordrecht: Kluwer), 487
Giacconi, R., Gursky, H., Paolini, F. R., and Rossi, B. B. (1962) PRL 9, 439
Giacconi, R., Gursky, H., Kellogg, E., Schreier, E., and Tananbaum, H. (1971), ApJ 167, L67Gilfanov, M., Revnivtsev, M., Sunyaev, R., and Churazov, E. (1998), A&A 338, L83
Glendenning, N. (2003), Compact Stars, (Springer-Verlag; 2nd edition)Grindlay, J., Gursky, H., Schnopper, H., Parsignault, D.R., Heise, J., Brinkman, A.C., and
Schrijver, J. (1976), ApJ 205, L127
Grimm, H.-J., Gilfanov, M., and Sunyaev, R. (2002), A&A 391, 923Grove, J. E., Johnson, W. N., Kroeger, R.A., McNaron-Brown, K., Skibo, J.G., and Phlips, B. F.
(1998), ApJ 500, 899
Hameury, J., Menou, K., Dubus, G., Lasota, J., and Hure, J. (1998), MNRAS 298, 1048Hands, S. (2001), Cont. Phys. 42, 209
Harlaftis, E.T. (2001), in Astrotomography, Indirect Imaging Methods in Observational
Astronomy, (Springer-Verlag) 359Hawley, J. F., and Krolik, J. H. (2001), ApJ 548, 348
Heindl, W. A. et al. (1999), ApJ 521, L49Heindl, W. A. et al. (2004), in X-ray Timing 2003: Rossi and Beyond, ed. P. Kaaret, F.K. Lamb,
and J. H. Swank (AIP), astro-ph/0403197Heinz, S., and Nowak, M.A. (2001), MNRAS 320, 249
Heyl, J. (2004), ApJ 600, 939Homer, L. et al. 2001, MNRAS 322, 827
Horne, K. (1985), MNRAS 213, 129———. (2003), in Astronomical Telescopes and Instrumentation, astro-ph/0301250
Igumenshchev, I. V., Abramowicz, M. A., and Narayan, R. (2000), ApJ 537, L27Igumenshchev, I. V., Narayan, R., and Abramowicz, M. A. (2003) ApJ 592, 1042
Illarionov, A. F., and Sunyaev, R. A. (1975), A&A 39, 185Jonker, P.G., Mendez, M., and van der Klis, M. (2002) MNRAS 336, L1
Kallman, T., Boroson, B., and Vrtilek, S. D. (1998), apj 502, 441Kato, M. (1983), PASJ 35, 33
Kato, S. (2001), PASJ 53, 1King, A. R., Kolb, U., and Burderi, L. (1996), ApJ 464, L127King, A.R., and Kolb, U. (1996), ApJ 481, 918
Kluzniak, W., Michelson, P., and Wagoner, R.V. (1990), ApJ 358, 538Kommers, J. M., Chakrabarty, D., and Lewin, W. H. G. (1998), ApJ 497, L33
Kong, A.K. H., Charles, P. A., and Kuulkers, E. 1998, New Astr. 3, 301Krolik, J. H., and Hawley, J. F. (2002), ApJ 573, 574
Kulkarni, S. R., and Narayan, R. (1998), ApJ 335, 755Kuulkers, E., van der Klis, M., and Vaughan, B. A. (1996), A&A 311, 197
Kuulkers, E., Wijnands, R., and van der Klis, M. (1999), MNRAS 308, 485
40 Accreting neutron stars and black holes:a decade of discoveries
Kuulkers, E., et al. (2003), A&A 399, 663Lasota, J.-P. (2000), A&A 360, 575
Laurent, P., and Titarchuk, L. (1999), ApJ 511, 289Lewin, W., van Paradijs, J., and Taam, R. in X-ray binaries
Liedahl, D.A. (1999), in X-ray Spectroscopy in Astrophysics, ed. J. van Paradijs and J. Bleeker,(Springer), 189
Liu, Q. Z., van Paradijs, J., and van den Heuvel, E. P. J. (2001), A&A 368, 1021Lorimer, D.R. (1995), MNRAS 274, 300
Lovelace, R. V. E., Romanova, M. M., and Bisnovatyi-Kogan, G. S. (1999), ApJ 514, 368Maloney, P. R., Begelman, M.C., and Pringle, J. E. (1996), ApJ 472, 582
Margon, B., and Anderson, S. F. (1999), ApJ 347, 448Markoff, S., Falcke, H., and Fender, R. (2001), A&A 372, L25
Markwardt, C.B., Smith, E, and Swank, J.H. (2003), ATEL, 122Markwardt, C.B., and Swank, J. H. (2003), IAUC 8144
Markwardt, C.B., Swank, J.H., Strohmayer, T.E., in’t Zand, J. J. M., and Marshall, F. E. (2002),ApJ 575, L21
Margon, B. (1984), ARA&A 22, 507
Marsh, T.R. (2000), in Lecture Notes in Physics: Astrotomography, ed. H. Boffin, D. Steeghs(Springer Verlag), astro-ph/0011020
Marshall, H. L., Canizares, C. R., and Schulz, N. S. (2002), ApJ 564, 941
McClintock, J. E., and Remillard, R.A. (1986), ApJ 308, 110McClintock, J. E. et al. (2001), ApJ 555, 477McKinney, J. C., and Gammie, C. F. (2002), ApJ 573, 728
Meszaros, P. (1992), High Energy Radiation from Magnetized Neutron Stars, (Univ. of ChicagoPress)
Miller, J. C., Shahbaz, T., and Nolan, L. A. (1998), MNRAS 294, L25
Miller, J. M. et al. (2002a), ApJ 570, L69Miller, J. M. et al. (2002b), ApJ 578, 348
Miller, M.C., Lamb, F. K., and Psaltis, D. (1998), ApJ 508, 791Mirabel, I. F., and Rodriguez, L. F. (1994), Nature 371, 46
Mirabel, I. F., Bandyopadhyay, R., Charles, P. A., Shahbaz, T., and Rodriguez, L. F. (1997), ApJ
477, L45Mirabel, I. F., et al. (1998), A&A 330, L9
Muno, M., Chakrabarty, D., Galloway, D., and Psaltis, D. (2002a), ApJ 580, 1048Muno, M., Ozel, F., and Chakrabarty, D. (2002b), ApJ 581, 550Narayan, R., Garcia, M.R., and McClintock, J. E. (2001), in Proc. IX Marcel Grossmann Meeting
eds. V. Gurzadyan, R. Jantzen and R. Ruffini (World Scientific: Singapore)Narayan, R., and Heyl, J. (2002), ApJ 575, L139Narayan, R., Igumenshchev, I. V., and Abramowicz, M. A. (2000), ApJ 539, 798
Narayan, R., McClintock, J. E., and Yi, I. (1996), ApJ 457, 821Narayan, R., and Yi, I. (1994), ApJ 428, L13
Narayan, R., Yi, I., and Mahadevan, R. (1995), Nature 374, 623Nath, N. R., Strohmayer, T. E., and Swank, J.H. (2002), ApJ 564, 353
Nelson, R. et al. (1998), ApJ 488, L117Nobili, L., Turolla, R., and Lapidus, I. (1994), ApJ 433, 276
O’Brien, K., Horne, K., Hynes, R. I., Chen, W., Haswell, C. A., and Still, M. D. (2002), MNRAS
334, 426Orosz, J.A. (2002) in A Massive Star Odyssey, from Main Sequence to Supernova, ed. K.A. van
der Hucht, A. Herraro, and C. Esteban (ASP: San Fransisco)
Orosz, J.A. and Kuulkers, E. (1999), MNRAS 305 132Paul, B., Kitamoto, S., and Makino, F. (2000), ApJ 528, 410Parmar, A. N., White, N. E., Stella, L., Izzo, C., and Ferri, P. (1989), ApJ 338, 359
Piran, T. (1978), ApJ 221, 652Priedhorsky, W.C., Terrell, J., and Holt, S. S. (1983), ApJ 270, 233
Priedhorsky, W.C., and Terrell, J. (1983), ApJ 273, 709———. (1984), ApJ 284, L17
Pringle, J. E. (1981), AR&AA 19, 137
1.4 Accretion-powered X-ray sources in the 21st century 41
———. (1996), MNRAS 281, 357Psaltis, D. (2004), in X-ray Timing 2003: Rossi and Beyond, ed. P. Kaaret, F.K. Lamb, and J. H.
Swank (AIP), astro-ph/0402213
Psaltis, D., and Chakrabarty, D. (1998), ApJ 521, 332Psaltis, D., Belloni, T., and van der Klis, M. (1999), ApJ 520, 262
Psaltis, D., and Lamb, F. K. (1998), in Neutron Stars and Pulsars, ed. N. Shibazaki, N. Kawai, S.Shibata, and T. Kifune (Tokyo: Universal Academy Press), 179
Psaltis, D., and Norman, C. (1999), astro-ph/0001391
Quataert, E., and Gruzinov, A. (1999), ApJ 520, 248Radhakrishnan, V., and Shrinivasan, G. (1982), Curr. Sci. 51, 1096Rhoades, C.E., and Ruffini, R. (1974), PRL 32, 324
Rutledge, R.E., Bildsten, L., Brown, E. F., Pavlov, G. G., and Zavlin, V.E. (2001), ApJ 551, 921Sano, T., Inutsuka, S., Turner, N. J., and Stone, J. M, (2004), ApJ 605, 321
Schatz, H., Bildsten, L., Cumming, A., and Wiescher, M. (2000), ApJ 524, 1014Schulz, N. S., Chakrabarty, D., Marshall, H. L., Canizares, C. R., Lee, J. C., and Houck, J.
(2001), ApJ 563, 941
Shakura, N. I., and Sunyaev, R. A. (1973), A&A 24, 337Shapiro, S. L., Lightman, A. P., and Eardley, D. M. (1976), ApJ 204, 187
Shapiro, S. L., and Teukolsky, S. (1983), Black Holes, White Dwarfs and Neutron Stars,(Wiley-Interscience)
Shirakawa, A., and Lai, D. 2002, ApJ 565, 1134
Shirey, R.E., Bradt, H. V., Levine, A.M., and Morgan, E. H. (1996), ApJ 469, L21Smale, A.P., and Lochner, J. C. (1992), ApJ 395, 582
Spencer, R.E. (1979), Nature 282, 483Spitkovsky, A., Ushomirsky, G., and Levin, Y. (2002), ApJ 566, 1018
Stella, L., Vietri, M., and Morsink, S. M. (1999), ApJ 524, L63Stone, J. M., and Pringle, J. E. (2001), MNRAS 322, 461Stone, J. M., Pringle, J. E., and Begelman, M. C. (1999), MNRAS 310, 1002
Strohmayer, T.E. (2001a), ApJ 552, L49———. (2001b), ApJ 554, L169
Strohmayer, T.E, and Brown, E. F. (2002), ApJ 566, 1045Strohmayer, T.E. et al. (1996), ApJ 469, L9
Sunyaev, R., and Revnivtsev, M. (2000), A&A 358, 617Taam, R.E., and Sandquist, E. L. (2000), ARA&A 38, 113
Titarchuk, L., Lapidus, I., and Muslimov, A. (1998), ApJ 499, 315Torkelsson, U. (1998), MNRAS 298, L55
Truemper, J., Pietsch, W., Reppin, C., Voges, W., Staubert, R., and Kendziorra, E. (1978), ApJ
219, L105van der Klis, M. (1994), ApJS 92, 511
———. (2000), ARA&A 38, 717van der Klis, M. et al. (1996), ApJ 469, L1
van Kerkwijk, M.H., Chakrabarty, D., Pringle, J. E., and Wijers, R. A.M. J. (1998), ApJ 499, L27van Kerkwijk, M. H., van Paradijs, J., and Zuiderwijk, E. J. (1995), A&A 303, 497van Paradijs, J. (1996), ApJ 464, L139
van Paradijs, J., and McClintock J. (1995), in X-ray Binaries, eds. W. H.G. Lewin, J. vanParadijs, and E.P. J. van den Heuvel (Cambridge: University Press)
Vaughan, B. et al. (1994), ApJ 435, 362
Vrielmann, S. (2000), in Lecture Notes in Physics: Astrotomography, ed. H. Boffin, D. Steeghs(Springer Verlag), astro-ph/0012263
Vrtilek, S. D., Raymond, J. C., Garcia, M. R., Verbunt, F., Hasinger, G., and Kurster, M. (1990),A&A 235, 162
Wagoner, R.W. (1999), Phys. Rep. 311, 259Wang, Z. et al. (2001), ApJ 563, L61
Waters, L. B. F. M., and van Kerkwijk, M. H. (1989), A&A 223, 196White, N. E., and Holt, S. S. (1982), ApJ 257, 318
White, N.E., Nagase, F., and Parmar, A.N. (1995), in X-ray Binaries, eds. W. H.G. Lewin, J.van Paradijs, and E.P. J. van den Heuvel (Cambridge: University Press).
42 Accreting neutron stars and black holes:a decade of discoveries
White, N. E., and Swank, J. H. (1982), ApJ 253, L61Wijnands, R.A.D., Kuulkers, E., and Smale, A. P. (1996), ApJ 473, L45Wijnands, R., and van der Klis, M. (1998), Nature 394, 344Will, C.M. (2001), Living Rev. Relativity 4, cited on 15 Aug 2001
(http://www.livingreviews.org/Articles/Volume4/2001-4will/)Wilms, J. et al. (2001), MNRAS 320, 327Witten, E. (1984), PRD 30, 272.Wojdowski, P. S. et al. (1998), ApJ 502, 253Yi, I., and Wheeler, J. G. (1998), ApJ 498, 802Zhang, W., Morgan, E. H., Jahoda, K., Swank, J. H., Strohmayer, T.E., Jernigan, G., and Klein,
R.-I. (1996), ApJ 469, L29Zhang, W., Smale, A. P., Strohmayer, T. E., and Swank, J.H. (1998), ApJ 500, L171Zingale, M. et al. (2001), ApJS 133, 195